Advertising Inside AI Assistants: ChatGPT vs Gemini vs Copilot vs Perplexity (2026)

If you bought ads a year ago, you had a few reliable choices: search, social, and maybe retail media. In 2026, a new surface has arrived and it is fragmenting fast. People now spend hours a week inside AI assistants, describing their problems in full sentences, and a handful of those assistants have started selling that attention to advertisers.

But “AI assistant advertising” is not one market. Each assistant sits at a different stage, with different formats, different economics, and different odds of the program still existing in a year. This guide walks through where you can actually buy ads today across ChatGPT, Google Gemini, Microsoft Copilot, and Perplexity, what is still just talk, and how to get your creative ready for a future where your buyers are split across all of them.

The AI assistant ad landscape

The reason this matters now is scale. AI assistants have gone from novelty to daily habit in under two years. ChatGPT reports roughly 900 million weekly active users, Google’s Gemini app has crossed 900 million monthly users, Meta AI claims about a billion, and Microsoft Copilot reaches several hundred million across Windows, Edge, Microsoft 365, and Bing. When a surface that large starts monetizing, advertisers cannot afford to treat it as experimental.

The important nuance, and the one most “AI ads” headlines miss, is the difference between advertising inside a chatbot conversation and advertising next to AI-generated content. eMarketer estimates that more than 80% of AI advertising in 2026 will appear next to AI content, like a Search or Shopping ad beside a Google AI Overview, rather than within a chatbot reply. The pure in-conversation ad, the kind ChatGPT and Perplexity pioneered, is still the smaller and riskier slice of the market.

That distinction shapes where the money is flowing. Google and Microsoft can graft ads onto AI experiences using their mature auction systems and existing advertiser base. Standalone assistants have to build both the ad product and the demand from scratch, which is exactly why their programs are moving at such different speeds.

In other words: the category is tiny today and enormous tomorrow. The advertisers who learn each surface while it is cheap and uncrowded will have a structural head start when budgets catch up. The rest of this guide breaks down each assistant so you know where to spend attention first.

ChatGPT ads

ChatGPT is the assistant that turned in-conversation advertising from theory into a live, self-serve product. OpenAI switched on ads for US users on February 9, 2026, and opened self-serve buying through its Ads Manager on May 5, with no minimum spend. That is a remarkable trajectory: the program went from a reported $200,000-minimum, agency-gated alpha (routed through holding companies like WPP, Omnicom, Dentsu, and Publicis) to no-minimum self-serve in under three months, according to reporting from Martech and other trade outlets.

Format and audience

The unit is a single “Sponsored” card that appears below the AI’s answer, never inside it. OpenAI has been explicit that advertising does not influence the model’s response. The card is compact: a headline of 3 to 50 characters, a description up to 100 characters, a square image (256×256 minimum, 512×512 recommended), and a small brand favicon. Because the unit renders inline and small, tight copy matters more than elaborate art.

Crucially, ads only show to logged-in US adults on the Free and Go ($8 per month) tiers. Plus, Pro, Business, Enterprise, and Education subscribers do not see them. So while ChatGPT’s total reach is huge, the addressable ad inventory is narrower than the headline user count suggests, and it skews toward consumers rather than heavy power users.

Targeting and pricing

ChatGPT does not sell keywords, demographics, or remarketing lists. Instead you supply context hints, short descriptions of the conversations where your product is relevant, and the system decides when to surface your card. Pricing has moved quickly: CPMs launched around $60 and have reportedly settled into a $25 to $60 range, with recommended CPC bids of $3 to $5. Early advertisers report conversion rates roughly 1.5× to 4× higher than Google for considered, research-heavy purchases, though attribution remains immature and those figures should be treated as directional.

For a full walkthrough of the format, the auction, and how to measure it, see our complete guide to ChatGPT ads. For how it stacks up against your existing channels, read ChatGPT ads vs Google ads vs Meta ads.

Perplexity ads

Perplexity is the cautionary tale of this category. It was one of the first AI companies to try advertising, and it is now one of the first to walk away. In November 2024 it launched sponsored follow-up questions, ad-generated prompts that appeared in the “Related Questions” area, alongside paid media positioned beside answers, with launch partners reportedly including Indeed, Whole Foods, Universal McCann, and PMG. Everything was clearly labeled, and the company insisted brands never wrote or edited the answers.

The experiment did not scale. By October 2025 Perplexity had stopped taking on new advertisers, and in February 2026 it stepped away from ads entirely, according to Financial Times reporting relayed by Search Engine Land, The Verge, and eMarketer. Its ad chief departed, reportedly fewer than 0.5% of applicant brands were ever admitted, and executives concluded that even labeled ads made users doubt the objectivity of every answer. As one executive put it, a user needs to believe they are getting the best possible answer, and ads put that belief at risk.

So who does Perplexity fit for right now? Practically speaking, no paid advertiser. With more than 100 million users and a reported 780 million monthly queries, it is a real audience, but there is currently no ad program, no self-serve access, and no managed buying. Executives have not ruled out returning to ads someday, but for now the only way into a Perplexity answer is organic citation, which makes it a target for answer-engine optimization (AEO) rather than a media buy. Treat this as reporting on a moving situation, not a permanent state.

Microsoft Copilot ads

Microsoft took the opposite approach to Perplexity: instead of building a bespoke chatbot ad product, it extended its existing ad machine into Copilot. Ads in Copilot are generated automatically from assets you already have in Microsoft Advertising, including Multimedia ads, Product ads, Responsive Search ads, and Performance Max campaigns. According to Microsoft’s own documentation, eligible campaigns are opted in by default and advertisers cannot opt out, though there is no guarantee any given ad will appear in a Copilot response.

Microsoft has published encouraging early numbers, reporting that ads in Copilot delivered click-through rates 69% stronger and conversion rates 76% higher than traditional search for lower-funnel ad types (Microsoft Advertising, 2024). As with any platform-reported figure, read it as a signal of direction rather than an audited benchmark.

The practical takeaway: if you already run Microsoft Advertising or Bing campaigns, you are very likely running Copilot ads without having done anything special. That makes Copilot the lowest-effort AI assistant placement available, but it also means you have limited control over the experience. Optimizing your existing feeds, assets, and Performance Max setup is the main lever you have to influence how you show up in Copilot conversations.

Google Gemini and AI Overviews

Google is the heavyweight, because it can put ads in front of AI users at a scale no standalone assistant can match. The Gemini app has grown from about 350 million monthly users in early 2025 to roughly 750 million by the end of the year and around 900 million by mid-2026, per Alphabet earnings and Google I/O disclosures. But the bigger story is AI Overviews, the Gemini-powered summaries inside Google Search, which reach an estimated two billion people a month.

Google is monetizing these experiences on two fronts. First, Search and Shopping ads now appear above, below, and within AI Overviews, labeled “Sponsored” and matched to both the query and the content of the AI answer. Second, at Marketing Live 2026 Google introduced Gemini-powered formats for AI Mode, its fully conversational search interface, including Conversational Discovery Ads, Highlighted Answers, AI-Powered Shopping Ads, and a Business Agent for Leads that lets users chat with a brand’s agent inside the ad unit.

The mechanics echo the other platforms: there is no separate “AI campaign” to create. Ads become eligible automatically if you run Performance Max, Shopping, or Search with AI-powered matching like broad match or AI Max for Search. Targeting keys off the full conversation context, not just the keyword. Google itself notes that ads in AI Overviews are still early and that reporting is evolving, so measure carefully and expect the rules to keep shifting.

Meta AI and others

Meta AI has the reach, about a billion monthly users across WhatsApp, Instagram, Messenger, and Facebook, but not yet the ads. Mark Zuckerberg has repeatedly said the company sees “a large opportunity to show product recommendations or ads” inside Meta AI, plus a premium subscription, but that Meta wants to grow engagement first. On recent earnings calls he suggested Meta would spend roughly another year scaling before building out the ad business in earnest. In May 2026 Meta began testing paid Meta AI plans (Meta One Plus at $7.99 and Premium at $19.99) in a few small markets, a sign that subscriptions, not ads, are the near-term monetization test inside the assistant.

Among the other assistants, xAI’s Grok and DeepSeek have meaningful audiences but no established self-serve ad programs to speak of. And Anthropic’s Claude, with an estimated 40 to 60 million weekly users, is the deliberate outlier: the company has committed to keeping Claude ad-free, and even ran Super Bowl ads mocking in-assistant advertising. We covered that split in depth in AI advertising and trust: OpenAI vs Anthropic. For advertisers, the takeaway is simple: Claude is not an ad channel, and betting on that changing soon would be speculative.

Side-by-side comparison

Here is the whole landscape in one view. Reach figures are approximate and mix weekly and monthly metrics as reported by each company, so treat them as orders of magnitude rather than precise comparisons.

Read top to bottom, the pattern is clear. Google and Microsoft offer scale and near-zero setup because they bolt AI ads onto systems you already use. ChatGPT is the one true native, self-serve chatbot ad product. Perplexity and Meta AI are, for now, not buyable, and Claude never will be by design.

How to prepare for a multi-assistant future

The strategic problem is not choosing one assistant. It is that your buyers are already scattered across all of them, and the surfaces open on their own timelines. The team that waits for a single winner will spend the next two years reacting late to each launch. The team that is ready with on-format creative can turn a placement on the day it becomes available.

That 80% figure is the practical guide for sequencing. Most near-term AI ad value lives in the Google and Microsoft ecosystems you can already access, so make sure your Performance Max feeds, assets, and Microsoft Advertising campaigns are healthy first. Then layer on ChatGPT, the one native chatbot placement that is live and self-serve, to start learning the format while it is cheap. Keep Perplexity and Meta AI on a watch list and revisit as the reporting changes.

The real bottleneck is production. Every surface wants a different creative shape, and hand-building each one is where teams stall. Here is the same campaign expressed as the native assets each placement demands.

Consistency across those rows is what keeps a brand recognizable as a user moves from a Google AI Overview to a ChatGPT card to an Instagram feed. Getting there manually means five briefs, five design cycles, and five sets of revisions. That is the exact friction that leaves most teams strong on one channel and absent on the rest. For the architecture behind this shift, see LLM ad infrastructure explained and our guide on how to build an LLM advertising stack.

One tool for every assistant

This is exactly the problem Lapis was built to solve. You describe your product and offer once, and Lapis generates native ad creative for every AI-native placement and every traditional channel, ChatGPT, Google, Meta, Reddit, and LinkedIn, from a single prompt in under three minutes, each asset auto-sized to the target format. Instead of rebuilding a campaign five times, you produce it once and ship everywhere.

Lapis is the first and only AI ad platform ready for ChatGPT, which means you can start learning the newest native placement today rather than after your competitors. That readiness is backed by real traction: a Y Combinator F25 company, a 5.0 rating on G2, and more than 10,000 campaigns created on the platform. When the next assistant opens its ad product, you already have the creative pipeline to move first.

You can start on the free tier to test the workflow, and most growth teams land on the Pro plan at $599 per month for full multi-platform output, analytics, and Campaign Studio. In a market where each assistant turns on ads at its own pace, the winning capability is not picking the right platform, it is being able to show up natively on all of them the moment they open. Start with Lapis and turn one brief into ready-to-run ads for every AI assistant and channel at once.